cuda - Thrust device_malloc and device_new -

what advantages using thrust device_malloc instead of normal cudamalloc , device_new do?

for device_malloc seems reason use it's bit cleaner.

the device_new documentation says:

"device_new implements placement new operator types resident in device memory. device_new calls t's null constructor on array of objects in device memory. no memory allocated function."

which don't understand...

device_malloc returns proper type of object if plan on using thrust other things. there no reason use cudamalloc if using thrust. encapsulating cuda calls makes easier , cleaner. same thing goes c++ , stl containers versus c-style arrays , malloc.

for device_new, should read following line of documentation:

 template<typename t>  device_ptr<t> thrust::device_new (device_ptr< void > p, const size_t n = 1)  

p: device_ptr region of device memory construct 1 or many ts.

basically, function can used if memory has been allocated. default constructor called, , return device_pointer casted t's type.

on other hand, following method allocates memory , returns device_ptr<t>:

template<typename t > device_ptr<t> thrust::device_new (const size_t n = 1)